Method and system for determining impassable area of obstacle in unmanned driving scenario

Abstract

The present invention discloses a method and system for determining an impassable area of an obstacle in an unmanned driving scenario. The method includes: rasterizing and splitting the clustered obstacle point cloud; determining the main direction of the point cloud in each grid based on the least squares method; calculating the parameters of the bounding box of the point cloud in each grid based on the minimum envelope principle; the parameters of the bounding box include: the length, width, height and center coordinates of the bounding box; determining the impassable area of the obstacle according to the parameters of the bounding box and the main direction of the point cloud in each grid. This method improves the deficiencies that when the obstacle point cloud is used as a bounding box to generate the impassable area, the shape of the bounding box is likely to be too different from the shape of the real obstacle, and when the size of the bounding box is too large, too many passable areas are misjudged as impassable areas. The present invention can improve the accuracy and robustness of obstacle detection for unmanned vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of driverless technology, and particularly to a method and system for determining an impassable area of an obstacle in a driverless scenario. Background Art

[0002] With the rapid development of technology and the improvement of people's living standards, the average number of cars owned by each family increases year by year. However, at the same time, the casualties and property losses caused by traffic accidents also increase year by year, posing a great threat to people's lives and safety and the harmony and stability of society. Driverless technology is an extension of the functions of traditional cars, endowing them with a high level of intelligence, enabling them to accurately perceive the surrounding environment, reasonably plan paths, and safely and autonomously reach their destinations. The intelligentization and networking of cars are an inevitable trend in the development of cars and the driving force for the continuous vitality of the automotive industry.

[0003] The environmental perception of a vehicle is the basis for realizing driverless driving and also the basis for subsequent decision-making, path planning, and control. Only when the driverless vehicle can accurately perceive the surrounding environment, especially the position, size, and direction of obstacles, and calculate the corresponding impassable area, can a series of subsequent functions be realized. However, in the complex scenarios of driverless driving, the shapes and sizes of obstacles are strange and varied, and not every type of obstacle can be simply represented by a bounding box. If those irregular-shaped obstacles are represented by a bounding box, it is very likely that the area of the impassable area will be too large, which is obviously unreasonable, leading to unnecessary difficulties in the subsequent vehicle path planning and even making it impossible to plan a feasible path in the entire map.

[0004] In the prior art, the method of generating an impassable area from unmanned obstacle point clouds usually represents the obstacle point clouds with an external circumscribed cube perpendicular to the ground, without considering that when the obstacle is irregular or long and strip-shaped, the shape of a single external circumscribed rectangle is too different from the actual contour of the obstacle, which will lead to an overly large coverage range of the calculated impassable area and affect subsequent planning. The methods of generating a single external circumscribed rectangle from obstacle point clouds as the impassable area mainly include two types: AABB and OBB. Although the AABB method has a simple principle and fast calculation speed and is the calculation method for the impassable area adopted in the vast majority of literature, it is difficult to accurately describe the boundary of the obstacle when the placement direction of the obstacle is inclined to the coordinate axis. OBB is an oriented bounding box, and the direction of the bounding box is no longer parallel to the coordinate axis. At this time, it is necessary to first determine the direction of the bounding box. In theory, this method can find the smallest rectangle containing the point cloud cluster, but in practical applications, when the number of point clouds in a category increases, the number of sides of the polygon also increases sharply, and there is no general algorithm to solve the problem of the smallest bounding rectangle of a convex n-sided polygon. And because the relationship between points and lines needs to be repeatedly judged in the algorithm and there are many loop traversal operations, when the number of obstacles and the number of point clouds increase, the calculation time will be very long. In addition, although the point cloud obstacle detection based on deep learning can fit a bounding box that is more conforming to the object contour, the current point cloud neural network can only process several fixed types of obstacles such as vehicles, pedestrians, and bicycles. Limited by the training data set, other types of obstacles cannot be recognized currently, nor can the impassable area of long and strip-shaped or irregular obstacle point clouds be split. Summary of the Invention

[0005] In view of the above problems, the purpose of the present invention is to provide a method and system for determining an impassable area of an obstacle in an unmanned driving scenario.

[0006] To achieve the above purpose, the present invention provides the following solutions:

[0007] A method for determining an impassable area of an obstacle in an unmanned driving scenario includes:

[0008] Rasterize and split the clustered obstacle point clouds;

[0009] Determine the main direction of the point cloud in each grid based on the least squares method;

[0010] Calculate the parameters of the bounding box of the point cloud in each grid based on the minimum envelope principle; the parameters of the bounding box include: the length, width, height, and center coordinates of the bounding box;

[0011] Determine the impassable area of the obstacle according to the parameters of the bounding box in each grid and the main direction of the point cloud.

[0012] Optionally, before rasterizing and splitting the clustered obstacle point cloud, it further includes: determining whether the clustered obstacle point cloud needs to be split; specifically including:

[0013] Obtain the maximum and minimum values in the x direction of all point clouds;

[0014] Obtain the maximum and minimum values in the y direction of all point clouds;

[0015] Calculate the scale of the point cloud in the x direction based on the maximum and minimum values in the x direction;

[0016] Calculate the scale of the point cloud in the y direction based on the maximum and minimum values in the y direction;

[0017] When the scale of the point cloud in the x direction and / or the scale of the point cloud in the y direction exceeds the scale threshold, it is determined that the clustered obstacle point cloud needs to be split.

[0018] Optionally, the calculation formula for determining the main direction of the point cloud in each grid based on the least squares method is as follows:

[0019] angle = acrtan(k)

[0020]

[0021] where angle represents the main direction angle, k represents the slope, N represents the number of points in the point cloud, x represents the X-axis coordinate of the point in the point cloud, and y represents the Y-axis coordinate of the point in the point cloud,

[0022] Optionally, the calculation process of the width of the bounding box is as follows:

[0023] Calculate the distance d from each point to the line y = kx w , and determine the maximum value d wmax and the minimum value d wmin ;

[0024] Calculate the width width of the bounding box according to the formula width = d wmax -d wmin Calculate the width width of the bounding box.

[0025] Optionally, the calculation process of the length of the bounding box is as follows:

[0026] Calculate the distance d from each point to the line and determine the maximum value d l and the minimum value d lmax and the minimum value d lmin ;

[0027] Calculate the length length of the bounding box according to the formula length = d lmax -d lmin Calculate the length length of the bounding box.

[0028] Optionally, the height of the bounding box is calculated as follows:

[0029] Obtain the maximum value z in the Z direction of all the point clouds in the grid max and the minimum value z min ;

[0030] According to the formula height = z max - z min Calculate the height heiht of the bounding box.

[0031] Optionally, the center coordinates of the bounding box are calculated as follows:

[0032] Calculate the two symmetry axes of the bounding box;

[0033] Determine the center X coordinate and center Y coordinate of the bounding box according to the intersection point of the two symmetry axes;

[0034] According to the maximum value z in the Z direction of the point cloud max and the minimum value z min Determine the center Z coordinate of the bounding box.

[0035] The present invention provides a system for determining an impassable area of an obstacle in an unmanned driving scenario, which is characterized by comprising:

[0036] A splitting module, configured to perform raster splitting on the clustered obstacle point cloud;

[0037] A main direction determination module, configured to determine the main direction of the point cloud in each grid based on the least squares method;

[0038] A parameter calculation module, configured to calculate the parameters of the bounding box of the point cloud in each grid based on the minimum envelope principle; the parameters of the bounding box include: the length, width, height and center coordinates of the bounding box;

[0039] An impassable area determination module of the obstacle, configured to determine the impassable area of the obstacle according to the parameters of the bounding box and the main direction of the point cloud in each grid.

[0040] According to the specific embodiments provided by the present invention, the following technical effects are disclosed:

[0041] The present invention determines whether rasterization processing needs to be performed according to the scale of the clustered obstacle point cloud, and calculates the bounding box for each cluster of point clouds in each grid. All the new information of the bounding boxes is combined together as the impassable area of the entire obstacle. This method improves the deficiencies existing when generating the impassable area by taking the obstacle point cloud as a bounding box, such as the shape of the bounding box being too different from the shape of the real obstacle and the size of the bounding box being too large, resulting in too many passable areas being misjudged as impassable areas. The present invention can improve the accuracy and robustness of obstacle detection for unmanned vehicles BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0043] Figure 1 It is a flowchart of a method for determining an impassable area of an obstacle in an unmanned driving scenario according to an embodiment of the present invention;

[0044] Figure 2 It is a schematic diagram of a method for determining a rasterization judgment threshold. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0046] The present invention provides a method for determining an impassable area of an obstacle in an unmanned driving scenario: judging whether it needs to be rasterized and split according to the scale of the clustered obstacle point cloud. For the obstacles that need to be split, each point is divided into the grid it belongs to. For the point cloud clusters in the same grid, the least squares method is used to determine its direction, and then the length and width of the bounding rectangle in the main direction are calculated based on the minimum envelope principle, so as to determine the parameters of the bounding box of each grid. Traverse all grids, calculate the parameters of each bounding box in the same way, and finally calculate the impassable area of the obstacle according to all the bounding box parameters.

[0047] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0048] As Figure 1 shown, the method for determining an impassable area of an obstacle in an unmanned driving scenario provided by the present invention specifically includes the following steps:

[0049] Step 101: Rasterize and split the clustered obstacle point cloud.

[0050] First, judge whether it needs to split the point cloud according to the following rules: First, obtain the maximum and minimum values in the x direction and y direction of all point clouds, and record them as x max 、xmin , y max and y min , calculate the scales of the point cloud in the x - direction and y - direction, which are L and W respectively, where

[0051] L = x max - x min

[0052] W = y max - y min

[0053] Manually set a size threshold SIZE. When the value of L or W is greater than this threshold, it is considered that the object has a large scale and the obstacle point cloud needs to be split. The setting of this threshold should ensure that only the point clouds of large obstacles are split, and the point clouds of common vehicle and pedestrian - type obstacles do not need to be split. This threshold can be determined according to the size of common vehicles. As Figure 2 shown, considering a vehicle as a cube perpendicular to the ground, when the diagonal of the rectangle in the top - view is parallel to the X or Y axis of the lidar, this clustering has the maximum L or W size. For general vehicle types, this threshold can be selected as 6 to 8m.

[0054] For the point - cloud clustering that needs to be split, perform rasterized splitting of the point cloud:

[0055] ① Set a grid size GRID_SIZE, divide the obstacle area into several square areas, then the number of grids in the X and Y directions are respectively

[0056]

[0057]

[0058] where the floor() function is to round up;

[0059] ② Create a three - dimensional container, where the third and second dimensions are num X and num Y , used to store the index numbers of the point cloud in each grid for subsequent traversal of each grid;

[0060] ③ Loop through each point in the clustered point cloud, obtain its X - coordinate and Y - coordinate, denoted as x and y, calculate its coordinates (i, j) in the grid, and store the index number of the point cloud into the corresponding container according to its coordinates

[0061]

[0062]

[0063] Step 102: Determine the principal direction of the point cloud in each grid based on the least squares method.

[0064] Calculate the principal direction of the point cloud in the XOY plane according to the least squares method:

[0065]

[0066] angle = acrtan(k)

[0067] Step 103: Calculate the parameters of the bounding box of the point cloud in each grid based on the minimum envelope principle; the parameters of the bounding box include: the length, width, height, and center coordinates of the bounding box.

[0068] ② Calculation of the width of the bounding rectangle:

[0069] Define the side parallel to the principal direction as the length and the side perpendicular to the principal direction as the width. Calculate the distance d from each point to the line y = kx w , and the calculation result does not need to take the absolute value, with positive and negative signs, and the calculation formula is as follows:

[0070]

[0071] Obtain the maximum value d and the minimum value d of the distances from the points in the point cloud to this line wmax and record the corresponding points as A(x1, y1) and B(x2, y2) respectively. Then the width of the bounding rectangle is wmin width = d

[0072] - d wmax wmin

[0073] ③ Calculation of the length of the bounding rectangle:

[0074] Calculate the distance d from each point to the line and record the corresponding points as C(x3, y3) and D(x4, y4) respectively. Then the width of the bounding rectangle is l , and the calculation result does not need to take the absolute value, with positive and negative signs, and the calculation formula is as follows

[0075]

[0076] Obtain the maximum value d and the minimum value d of the distances from the points in the point cloud to this line max and record the corresponding points as C(x3, y3) and D(x4, y4) respectively. Then the width of the bounding rectangle is min length = d

[0077] - d lamx lmin

[0078] ④ Calculation of the center of the bounding rectangle:

[0079]

[0079] Based on the slope k calculated according to the above steps and the coordinates of the four edge points A, B, C, and D, the equations of the four sides can be calculated, which are respectively:

[0080] l A : y = kx + (-kx1 + y1) = kx + b1

[0081] l B : y = kx + (-kx2 + y2) = kx + b2

[0082]

[0083]

[0084] Among them, l A and l B are the equations of the two long sides, and l C and l D are the equations of the two short sides. According to these two sets of straight-line equations, the equations of the two symmetry axes of the bounding rectangle are calculated, which are respectively:

[0085]

[0086]

[0087] Calculating the intersection point of these two straight lines is the center coordinate of the bounding rectangle

[0088]

[0089] center Y = k * center X + b1

[0090] ⑤ Bounding box height and center Z coordinate calculation:

[0091] Obtain the maximum value z max and the minimum value z min

[0092] height = z max - z min

[0093] Step 104: Determine the impassable area of the obstacle according to the parameters of the bounding box in each grid and the main direction of the point cloud.

[0094] For the point cloud in each grid, the bounding box parameters can be calculated according to the above steps: center X 、center Y 、centr Z, length, width, height, and angle can be used to draw corresponding impassable areas according to the bounding box parameters of the entire category.

[0095] According to the scale of the clustered obstacle point cloud, the present invention determines whether rasterization processing is required, calculates the bounding box for each point cloud cluster in each grid separately, and combines all the new bounding box information as the impassable area of the entire obstacle. This method improves the deficiencies that occur when generating the impassable area with the obstacle point cloud as a single bounding box, where the shape of the bounding box is too different from the actual obstacle shape, and the size of the bounding box is too large, resulting in too many passable areas being misjudged as impassable areas.

[0096] The present invention provides a system for determining the impassable area of obstacles in an unmanned driving scenario, including:

[0097] A splitting module for rasterizing and splitting the clustered obstacle point cloud;

[0098] A main direction determination module for determining the main direction of the point cloud in each grid based on the least squares method;

[0099] A parameter calculation module for calculating the parameters of the bounding box of the point cloud in each grid based on the minimum envelope principle; the parameters of the bounding box include: the length, width, height, and center coordinates of the bounding box;

[0100] An obstacle impassable area determination module for determining the obstacle impassable area according to the parameters of the bounding box in each grid and the main direction of the point cloud

[0101] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0102] In this article, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for determining an impassable area of an obstacle in an unmanned driving scenario, characterized in that, Including: Rasterize and split the clustered obstacle point cloud; Determine the main direction of the point cloud in each grid based on the least squares method; Calculate the parameters of the bounding box of the point cloud in each grid based on the minimum envelope principle; The parameters of the bounding box include: the length, width, height and center coordinates of the bounding box; define the side parallel to the main direction as the length, and the side perpendicular to the main direction as the width; Determine the impassable area of the obstacle according to the parameters of the bounding box in each grid and the main direction of the point cloud.

2. The method for determining an impassable area of an obstacle in an unmanned driving scenario according to claim 1, wherein Before rasterizing and splitting the clustered obstacle point cloud, it also includes: judging whether the clustered obstacle point cloud needs to be split; specifically including: Obtain the maximum and minimum values in the x direction of all point clouds; Obtain the maximum and minimum values in the y direction of all point clouds; Calculate the scale of the point cloud in the x direction according to the maximum and minimum values in the x direction; Calculate the scale of the point cloud in the y direction according to the maximum and minimum values in the y direction; When the scale of the point cloud in the x direction and / or the scale of the point cloud in the y direction exceeds the scale threshold, it is determined that the clustered obstacle point cloud needs to be split.

3. The method for determining an impassable area of an obstacle in an unmanned driving scenario according to claim 1, wherein The calculation formula for determining the main direction of the point cloud in each grid based on the least squares method is as follows: angle = acrtan(k) Where, angle represents the main direction angle, k represents the slope, N represents the number of points in the point cloud, x represents the X-axis coordinate of the point in the point cloud, and y represents the Y-axis coordinate of the point in the point cloud.

4. The method for determining an impassable area of an obstacle in an unmanned driving scenario according to claim 3, wherein The calculation process of the width of the bounding box is as follows: Calculate the distance d from each point to the line y = kx w , and determine the maximum value d wmax and the minimum value d wmin ; Calculate the width of the bounding box according to the formula width = d wmax -d wmin ​ 5. The method for determining an impassable area of an obstacle in an unmanned driving scenario according to claim 3, wherein The calculation process of the length of the bounding box is as follows: Calculate the distance d from each point to the straight line and determine the maximum value d l and the minimum value d lmax ; lmin ; Calculate the length of the bounding box according to the formula length = d lmax - d lmin ​ 6. The method for determining an impassable area of an obstacle in an unmanned driving scenario according to claim 3, wherein The calculation process of the height of the bounding box is as follows: Obtain the maximum value z of all point clouds in the raster in the Z direction max and the minimum value z min ; Calculate the height of the bounding box according to the formula height = z max -z min ​ 7. The method for determining an impassable area of an obstacle in an unmanned driving scenario according to claim 6, wherein The calculation process of the center coordinates of the bounding box is as follows: Calculate the two symmetry axes of the bounding box; the symmetry axes are the diagonals of the bounding box; Determine the center X coordinate and center Y coordinate of the bounding box according to the intersection point of the two symmetry axes; Determine the center Z coordinate of the bounding box according to the maximum value z and the minimum value z of the point cloud in the Z direction. max and the minimum value z min of the point cloud in the Z direction.

8. An obstacle non-passable area determination system in an unmanned driving scenario, characterized in that, Including: A splitting module for rasterizing and splitting the clustered obstacle point cloud; A main direction determination module for determining the main direction of the point cloud in each grid based on the least squares method; A parameter calculation module for calculating the parameters of the bounding box of the point cloud in each grid based on the minimum envelope principle; The parameters of the bounding box include: the length, width, height and center coordinates of the bounding box; define the side parallel to the main direction as the length, and the side perpendicular to the main direction as the width; An impassable area determination module for the obstacle, which is used to determine the impassable area of the obstacle according to the parameters of the bounding box in each grid and the main direction of the point cloud.

Images